How-To Chat with a model in YoKI
The following instructions explain step by step how you can select one of the numerous language models in YoKI and interact with them.
Note: The URZ is constantly working to update and expand the models in YoKI. On this page, you can view a list of the models currently available.
Contents
Presentation of the models
YoKI currently comprises four language models with different areas of application:
- LLaMA 3.1 (70B): Meta's model is designed for argumentation and translation tasks. The model is particularly suitable for everyday use and general tasks. (Note: This model is set as the default.)
- DeepSeek-R1 Distill: This LLM is designed for everyday tasks and is particularly useful in environments where GPU memory is limited.
- Aya-23-8B: This model is particularly suitable for multilingual tasks. It has been trained on 23 languages, including Arabic, Indonesian, Japanese, Romanian, Russian, German, and English.
- Qwen2.5-Coder-14B: Qwen is focused on generating, improving, and extending code. It is capable of writing, debugging, and commenting on code in Python, C++, and other languages.
Select a model
1. Open the YoKI web interface.
2. In the chat window, click on the model that is currently set in the upper left corner. (Note: LLaMA 3.1 (70B) is set by default.)
3. Select the model you want to interact with.
Note: You can filter for a specific model by name under “Search models...”. You can recognize the selected model by the check mark to the right of the model description.
4. Write a message in the message window and send it.
Tips for prompts
What is a prompt?
A prompt is an instruction sent to a large language model or AI system to obtain a specific action. It is our human input (e.g., text or voice input) that an AI system processes into an output (e.g., an image or text).
What do effective prompts look like?
There are a few tips that should be kept in mind when it comes to prompt engineering:
1. Defining the context and a role
- Instead of simply asking a question, it makes sense to also assign a role and situational context to an AI system.
-
Example:
- ❌ Not useful: “Write a blog article about microservices.”
- ✅ Useful: “As an experienced software architect with 15 years of experience in building distributed systems, write a technical blog article about patterns for microservices architecture.”
2. Clearly defined goals and limitations
- Specify your desired result and any restrictions. This will enable the AI system to work more closely on your problem or question.
-
Example:
- ❌ Not useful: “Help me with my Python code.”
- ✅ Useful: “Review this Python function for performance improvement. Focus on reducing memory usage and improving runtime complexity.”
3. Definition of the output format
- Define exactly how you want the information to be structured.
-
Example
- ❌ Not useful: “Tell me about different types of databases.”
- ✅ Useful: „Compare NoSQL and SQL databases based on the following aspects: use cases, performance characteristics, and scalability."
In general, your prompts should be detailed and explicitly state what you expect from the AI system.